This is the base class for a number of render attributes (other than transform) that may be set on scene graph nodes to control the appearance of geometry. This includes
RenderAttrib represents render attributes that always propagate down to the leaves without regard to the particular node they are assigned to. A RenderAttrib will have the same effect on a leaf node whether it is assigned to the graph at the leaf or several nodes above. This is different from
RenderEffect, which represents a particular render property that is applied immediately to the node on which it is encountered, like billboarding or decaling.
You should not attempt to create or modify a RenderAttrib directly; instead, use the make() method of the appropriate kind of attrib you want. This will allocate and return a new RenderAttrib of the appropriate type, and it may share pointers if possible. Do not modify the new RenderAttrib if you wish to change its properties; instead, create a new one.
intentionally defined to match D3DCMPFUNC
enumerator M_none = 0
alpha-test disabled (always-draw)
enumerator M_never = 1
enumerator M_less = 2
incoming < reference_alpha
enumerator M_equal = 3
incoming == reference_alpha
enumerator M_less_equal = 4
incoming <= reference_alpha
enumerator M_greater = 5
incoming > reference_alpha
enumerator M_not_equal = 6
incoming != reference_alpha
enumerator M_greater_equal = 7
incoming >= reference_alpha
enumerator M_always = 8
- enumerator M_none = 0
enumerator M_off = 0
enumerator M_eye_sphere_map = 1
Sphere maps are classic static reflection maps. They are supported on just about any hardware, and require a precomputed 360-degree fisheye image. Sphere maps only make sense in eye coordinate space.
enumerator M_world_cube_map = 2
/* * Cube maps are a modern improvement on the sphere map; they don’t suffer * from any polar singularities, but they require six texture images. They * can also be generated dynamically for real-time reflections (see * GraphicsOutput::make_cube_map()). Typically, a statically-generated cube * map will be in eye space, while a dynamically-generated map will be in * world space. Cube mapping is not supported on all hardware. */
enumerator M_eye_cube_map = 3
enumerator M_world_normal = 4
Normal maps are most useful for applying diffuse lighting effects via a pregenerated cube map.
enumerator M_eye_normal = 5
enumerator M_world_position = 6
Position maps convert XYZ coordinates directly to texture coordinates. This is particularly useful for implementing projective texturing (see NodePath::project_texture()).
enumerator M_unused = 7
formerly M_object_position, now deprecated.
enumerator M_eye_position = 8
enumerator M_point_sprite = 9
/* * With M_point_sprite, texture coordinates will be generated for large points * in the range (0,0) - (1,1) from upper-left to lower-right across the * point’s face. Without this, each point will have just a single uniform * texture coordinate value across its face. Unfortunately, the generated * texture coordinates are inverted (upside-down) from Panda’s usual * convention, but this is what the graphics card manufacturers decided to * use. You could use a texture matrix to re-invert the texture, but that * will probably force the sprites’ vertices to be computed in the CPU. You’ll * have to paint your textures upside-down if you want true hardware sprites. */
enumerator M_unused2 = 10
M_light_vector generated special 3-d texture coordinates that represented the vector to a particular Light in the scene graph, expressed in each vertex’s tangent space. This has now been removed. We need to reserve the slot in the enum, though, to make sure the following enum value still has the same value.
enumerator M_constant = 11
M_constant generates the same fixed texture coordinates at each vertex. Not terribly useful, of course, except for certain special effects involving moving a flat color over an object.
- enumerator M_off = 0
int compare_to(RenderAttrib const &other) const
Provides an arbitrary ordering among all unique
RenderAttribs, so we can store the essentially different ones in a big set and throw away the rest.
ConstPointerTo<RenderAttrib> compose(RenderAttrib const *other) const
Returns a new
RenderAttribobject that represents the composition of this attrib with the other attrib. In most cases, this is the same as the other attrib; a compose b produces b. Some kinds of attributes, like a TextureTransform, for instance, might produce a new result: a compose b produces c.
static int garbage_collect(void)
Performs a garbage-collection cycle. This is called automatically from
RenderState::garbage_collect(); see that method for more information.
std::size_t get_hash(void) const
Returns a suitable hash value for phash_map.
static int get_num_attribs(void)
Returns the total number of unique
RenderAttribobjects allocated in the world. This will go up and down during normal operations.
virtual int get_slot(void) const = 0
ConstPointerTo<RenderAttrib> get_unique(void) const
Returns the pointer to the unique
RenderAttribin the cache that is equivalent to this one. This may be the same pointer as this object, or it may be a different pointer; but it will be an equivalent object, and it will be a shared pointer. This may be called from time to time to improve cache benefits.
ConstPointerTo<RenderAttrib> invert_compose(RenderAttrib const *other) const
Returns a new
RenderAttribobject that represents the composition of the inverse of this attrib with the other attrib. In most cases, this is the same as the other attrib; !a compose b produces b. Some kinds of attributes, like a TextureTransform, for instance, might produce a new result: !a compose b produces c.
This is similar to
compose()except that the source attrib is inverted first. This is used to compute the relative attribute for one node as viewed from some other node, which is especially useful for transform-type attributes.
static void list_attribs(std::ostream &out)
Lists all of the
RenderAttribsin the cache to the output stream, one per line. This can be quite a lot of output if the cache is large, so be prepared.
virtual bool lower_attrib_can_override(void) const
This should return false if a
RenderAttribon a higher node will compose into a
RenderAttribon a lower node that has a higher override value, or true if the lower
RenderAttribwill completely replace the state.
The default behavior is false: normally, a
RenderAttribin the graph cannot completely override a
RenderAttribabove it, regardless of its override value–instead, the two attribs are composed. But for some kinds of
RenderAttribs, it is useful to allow this kind of override.
This method only handles the one special case of a lower
RenderAttribwith a higher override value. If the higher
RenderAttribhas a higher override value, it always completely overrides. And if both
RenderAttribshave the same override value, they are always composed.
virtual void output(std::ostream &out) const
static bool validate_attribs(void)
Ensures that the cache is still stored in sorted order. Returns true if so, false if there is a problem (which implies someone has modified one of the supposedly-const
virtual void write(std::ostream &out, int indent_level) const
- enum PandaCompareFunc